Oswald

Abstract

The well-known Smith-Waterman (SW) algorithm is a high-sensitivity method for local sequence alignments. Unfortunately, SW has quadratic time complexity, which makes this algorithm computationally demanding for large protein databases. In this paper, we present OSWALD, a portable, fully functional and general implementation to accelerate SW database searches in heterogeneous platforms based on Altera's FPGA. OSWALD exploits OpenMP multithreading and SIMD computing through SSE and AVX2 extensions on the host while it takes advantage of pipeline and vectorial parallelism by way of OpenCL on the FPGAs. Performance evaluations on two different heterogeneous architectures with real amino acids datasets show that OSWALD is competitive in comparison with other top-performing SW implementations reaching up to 442 GCUPS peak with the best GCUPS/watts ratio.

SWIFOLD: Smith-Waterman implementation on FPGA with OpenCL for long DNA sequences

“…2). It is important to mention that, although in the OSWALD implementation [10] Intel/Altera OpenCL channels are used to communicate these data, the use of this technique is not feasible in the context of DNA with millions of nucleotide bases involved, since its size would exceed by far the channel resources available. We should point out that although the use of these buffers could double memory consumption, it is by far compensated on speedup terms.…”

“…Most of them correspond to protein alignment, and are parallelized on High-Performance Computing (HPC) architectures [7] and emerging architectures [8–10]. For very long sequences, such as with DNA, the number of works is significantly lower.…”

SWIMM 2.0: Enhanced Smith–Waterman on Intel’s Multicore and Manycore Architectures Based on AVX-512 Vector Extensions

“…The first three servers were used to evaluate SWIMM 2.0 and the other SIMD-based alternatives, while the rest were employed to perform a comparison with GPUs and FPGAs. The performance was evaluated by carrying out similar experiments to those in previous works [19,14,11,7]. We have evaluated SWIMM 2.0 by searching 20 query protein sequences against three well-known databases of different size:…”

“…This version computes using 32-bit integer data but allow us to include newer GPUs in the analysis. For FPGA accelerators, we have chosen the OS-WALD package [19] in its hybrid configuration because it offers a satisfactory performance-power tradeoff. Table 2 presents power efficiency ratios considering the GCUPS peak performance and the Thermal Design Power (TDP) of each platform.…”

“…Moreover, the recently released SWhybrid framework offers the possibility of combining CPUs, GPUs and Xeon Phi's [15]. Finally, there are also studies that use Field Programmable Gate Arrays (FPGAs) as accelerators, such as linear systolic array implementations for Xilinx Virtex FPGAs [16,17]; custom instructions [18] and the novel OpenCL paradigm on Intel-Altera's FPGAs [19].…”